Current status of ld-decode / vhs-decode (true "backup" of RF signals)

[harrypm]

There is not really much point in a A/B test if the FM RF capture config and files are not there alongside and decode version/args stated. (lets not make more crafters examples lol)

Should also be noted since --0ire_adjust was implimented black levels are now automatically dead accurate on the majorty of tapes without segment tweaks required at all.

tbc-video-export is also not a dumb script it has 720x480 / 720x576 and IMX 720x512 / 720x608 alongside the 4fsc luma output mode profiles for all codecs if you tell it to use them.

But the perfectly clean headswitch line clearly gives away what one is decode!

[dellsam34]

The first video and I assume that's the one captured from the VCR video output based on the menu at the beggining looks much better, the artifacts like hallows and ringing shown in the second video are not in the VCR version. This could mean two things: One, the RF pickup at the hardware level sucks and it's not up to the task since it's creating such artifacts. Two, If the artifacts are created by the tape-decode software and not the RF hardware that means the VCR processing proven to be better than the tape-decode processing. In either case the tape-decode is still way behind a conventional playback craeted 4 decades ago.

Edit:
The excessive head switch in the VCR version can be adjusted by either reclocking the head drum stator or via a potentionmeter in the main board.
I also believe the VCR should be in dubbing mode to disable picture processing and gain more sharpness, I also prefer the 629TBC to be off since not all VCRs have it.

[WaxCyl]

Thanks for all your work Brad. The chroma appears to be offset for the VHS-decode sample putting it at a slight disadvantage here. This should be corrected before comparison. The conventional sample has more ghosting and ringing so is already oversharpened. I think it would make a fair comparison to give the VHS-decode sample some modern software sharpening as it doesn't currently have that benefit. I think it is important to match black and white levels as more contrast will give a visual boost to favour the one with more contrast, especially in text!

[dellsam34]

This should be corrected before comparison. The conventional sample has more ghosting and ringing so is already oversharpened. I think it would make a fair comparison to give the VHS-decode sample some modern software sharpening as it doesn't currently have that benefit. I think it is important to match black and white levels as more contrast will give a visual boost to favour the one with more contrast, especially in text!

That wouldn't be a comparaison if you start tipping the scale, Raw capture is what we asked for not processed ones, As to ghosting and ringing, I believe you reversed the samples, The vhs-decode sample that has those artifacts.

[Sharc]

Comparison script in post#1079: Shouldn't we set converttoYV24(interlaced=true) to process the chroma correctly?

[Brad]

Comparison script in post#1079: Shouldn't we set converttoYV24(interlaced=true) to process the chroma correctly?

Doesn't make a difference when the source is 4:2:2, so I saved the 15 bytes (times two). http://avisynth.nl/index.php/Convert#Interlaced

I'll address the other points later. Thanks all for your feedback.

[WaxCyl]

Look at the LHS image edge of the 2nd (robot) image to see all the edge artifacts of the conventional capture.

That wouldn't be a comparaison if you start tipping the scale, Raw capture is what we asked for not processed ones, As to ghosting and ringing, I believe you reversed the samples, The vhs-decode sample that has those artifacts.

Fair enough if that's how you want to compare. I just can't understand how you can make an image quality assessment on an almost raw signal.The inherent advantage of the RF captured video is that it has not been processed by a "90s" VCR which has been over-sharpened to the max where any further post sharpening looks awful, whereas RF captured can be enhanced far more (an inherent advantage). I make my own comparisons by determining the maximum amount each video can be enhanced - RF then wins every time.

[lollo]

The inherent advantage of the RF captured video is that it has not been processed by a "90s" VCR which has been over-sharpened to the max where any further post sharpening looks awful

This is not true. All my "conventional" captures benefit from a sharpening in post processing. But maybe you are referring to the "bad" "conventional" captures often used for comparison.

whereas RF captured can be enhanced far more (an inherent advantage).

According to the samples posted by Brad not really. The vhs-decode is already sharpened compared to the "conventional" capture, so actually is the opposite. I will try to put some example later...

I make my own comparisons by determining the maximum amount each video can be enhanced - RF then wins every time.

Where are your samples? Please provide a fair comparison, and not the usual useless bad "conventional" captures.

[Brad]

^ @lollo is "conventional" a bad word?

There is not really much point in a A/B test if the FM RF capture config and files are not there alongside and decode version/args stated.
[...]
Should also be noted since --0ire_adjust was implimented black levels are now automatically dead accurate on the majorty of tapes without segment tweaks required at all.

In my experience, they're always very slightly brighter than Y=16.

Here are the arguments used:

vhs-decode -t3 --no_resample --ire0_adjust INFILE OUTFILE

Version is... the current one. I'm on the wrong computer to check right now.

I've added the RF sample to the post, as 3-part ZIP created by 7zip. I didn't upload it previously due to size and horrendously slow upload speed to this site (190KB/sec).

tbc-video-export is also not a dumb script it has 720x480 / 720x576

Where are those? The Wiki doesn't list resolutions, except IMX.

The excessive head switch in the VCR version can be adjusted by either reclocking the head drum stator or via a potentiometer in the main board.

On my US model that uses the same remote codes (Mitsubishi HS-HD2000U) it's an automatic adjustment by shorting TP5A and TP5B. Without any way to get the service manual from Japan, I'm afraid of wrecking this VCR by attempting anything.

I also believe the VCR should be in dubbing mode to disable picture processing and gain more sharpness, I also prefer the 629TBC to be off since not all VCRs have it.

I'm surprised you want to disable TBC, but sure, the next comparison will be Dubbing mode minus 629TBC.

This should be corrected before comparison. The conventional sample has more ghosting and ringing so is already oversharpened. I think it would make a fair comparison to give the VHS-decode sample some modern software sharpening as it doesn't currently have that benefit. I think it is important to match black and white levels as more contrast will give a visual boost to favour the one with more contrast, especially in text!

That wouldn't be a comparison if you start tipping the scale, Raw capture is what we asked for not processed ones

Indeed, that's why I shared the "raw" decode without manual adjustments but paired it with an Avisynth comparison script. If anyone has a way to process the samples further to improve quality, please share.

[lollo]

^ @lollo is "conventional" a bad word?

Not at all. I used that word just to differentiate the 2 workflows

[lollo]

Some more analysis.

Levels are ok for both captures

ATI USB 600:


VHS-Decode:


As noticed yesterday, on the text portion of the video ATI USB 600 capture is better: https://imgsli.com/MzUwNjYx

on the CGI part of the video VHS-Decode seems more defined, unless I confuse details with artifacts: https://imgsli.com/MzUwNjYz

After applying some lite sharpening on the ATI USB 600 capture: https://imgsli.com/MzUwNjY0

the text portion of the video ATI USB 600 capture is (obviously) even better: https://imgsli.com/MzUwNjY2

while on the CGI part of the video I still see a more detailed image on VHS-Decode capture, in the background wall and in the tecture of the character, unless I confuse details with artifacts: https://imgsli.com/MzUwNjY3

And finally, applying some sharpening on the VHS-Decode (not much improvement here, strange, I have to investigate): https://imgsli.com/MzUwNjc1

the capture is still not able to make it comparable with ATI USB 600 capture for the text portion: https://imgsli.com/MzUwNjc3

while on the CGI portion the improvement is not evident: https://imgsli.com/MzUwNjc4

To do: try some denoise on the CGI portion of the VHS-Decode capture.

Conclusion: watching the text portion of the video there is a clear winner, the ATI USB 600 capture, while for the animation part I am not sure of what is better, because VHS-Decode seems more detailed, even if I prefer the look of the ATI USB 600 capture.
Apparently, the AviSynth sharpening on the VHS-Decode capture is less effective than in the ATI 800 USB capture.

edit: I forgot to mention that the sharpening I applied is a simple LSFmod(defaults="slow")

[dellsam34]

I'm surprised you want to disable TBC, but sure, the next comparison will be Dubbing mode minus 629TBC.

I was not aware that "Dubbing" disables TBC, I though it behaves like "Edit" for JVC models.

[dellsam34]

on the CGI part of the video VHS-Decode seems more defined, unless I confuse details with artifacts: https://imgsli.com/MzUwNjYz

I see both samples have the same sharpness, The VHS-Decode (or tape-decode as they call it now since it captures different formats not just VHS) has an extra layer of noise which makes it look sharper but the actual video details are not.

I started to think that the RF hardware is not up to the task, not the decode part of it, Judging by the horizontal resolution of the white text on the blue background, it seems to me that there is not enough bandwidth to capture more details of each scan line.

[Sharc]

The ATI600 (TFF) has a tint shift between even and odd fields (clean<->purple).
The VHSDecode (BFF) seems to have signal reflexions (text). Caused by an impedance mismatch somewhere?

[lollo]

I started to think that the RF hardware is not up to the task, not the decode part of it, Judging by the horizontal resolution of the white text on the blue background, it seems to me that there is not enough bandwidth to capture more details of each scan line.

Well, based on the text portion of the video, which should be easier to judge, being an objective reference, I also think that the "conventional" workflow produces better results.

On my side I had great expectations on vhs-decode, not confirmed so far. On the other hand, when I compare my "conventional" capture with the digital sources broadcasted and used as reference, I see same quality, so I think "nothing can be better than that".

I would also like to see a comparison with real footage (actors) and not CGI.

For sure we need more tests and samples to give a final judgement.

[aramkolt]

Probably not directly related, but I'll comment for any high end U-Matic users out there. On the BVU9x0 series, if you use the actual dub outputs (they use those fun 7 pin DIN connectors) the internal TBC does, in fact, get disabled. This is clearly the case as the "TBC" light on the front turns off when you do it. This is quite a shame because the TBCs in those are pretty robust. The next best thing would be to grab S-Video off of the TBC card, but alas, the first thing the TBC does is combine luma and chroma into composite before putting the composite signal into a digital 16 line buffer, so if that internal TBC is used, then you're kind of stuck with composite output from that.

I was generally not as impressed with the DD over a traditional capture with my very limited testing off of an AG1980 that I fully refurbished, but that's been a year or two ago now. I did want to try it with U-Matic, but I haven't been able to get it working with my Mac reliably, though others say it works fine, so could just be me I guess. I'll get around to dragging out my windows tower PC eventually and see if it fares any better on that in terms of getting it to capture the RF.

The other thing that may or may not have been mentioned is that the DD puts a 50 ohm load to ground on the RF test point, so depending on your player, the regular outputs of the machine might not be usable during an RF capture which will be a hindrance to anyone wanting to simultaneously capture RF and traditionally. There's a separate device/PCB that can be built to act as a buffer so that a load isn't directly applied to the RF, but I forget the name of it at the moment. Quite possible that others have already mentioned this in the thread, but this is such a long thread that it is hard to really "search" that. May not be the case with most VCRs, but I did see extremely distorted video immediately on the U-Matic machine when I connected the DD to the test point.

[harrypm]

There is not really much point in a A/B test if the FM RF capture config and files are not there alongside and decode version/args stated.
[...]
Should also be noted since --0ire_adjust was implimented black levels are now automatically dead accurate on the majorty of tapes without segment tweaks required at all.

In my experience, they're always very slightly brighter than Y=16.

Here are the arguments used:

vhs-decode -t 3 --no_resample --ire0_adjust INFILE OUTFILE

Version is... the current one. I'm on the wrong computer to check right now.

tbc-video-export is also not a dumb script it has 720x480 / 720x576

Where are those? The Wiki doesn't list resolutions, except IMX.

For example tbc-video-export can use any FFmpeg filters you wish as that is what it's profiles are based around, so you can just make a profile for a config you like.

tbc-video-export  input.tbc --append-video-filter "scale=720x480:flags=lanczos,setdar=4/3"

It's only a forced profile for x264_web & x265_web and D10 non-imx profiles iirc, as most people will be tossing it into Resolve/StaxRip or whatever flavtor of standard post processing pipelines.

The other thing that may or may not have been mentioned is that the DD puts a 50 ohm load to ground on the RF test point, so depending on your player, the regular outputs of the machine might not be usable during an RF capture which will be a hindrance to anyone wanting to simultaneously capture RF and traditionally. There's a separate device/PCB that can be built to act as a buffer so that a load isn't directly applied to the RF, but I forget the name of it at the moment. Quite possible that others have already mentioned this in the thread, but this is such a long thread that it is hard to really "search" that. May not be the case with most VCRs, but I did see extremely distorted video immediately on the U-Matic machine when I connected the DD to the test point.

The DdD is well mentioned in all the current wiki docs key one being the "Workflow Guide" that it is a "LaserDisc" focused device, this is why the CX Cards + Clockgen is the entry gold standard for users who care about VHS/Betamax/Umatic and outher tape formats and not single channel formats like LD/Video8/Hi8/BetaMax NTSC.

(CX Cards now work on Windows, and the decode suite runs fluidly on Windows, if you want 100% always updated Mac support then feel free to buy some M1/X86 systems for the dev group, I never ignored MacOS support, and the new DdD app rewrite does support it too.)

The amplifyer is the duel channel ADA4857, which has a dedicated wiki page and is in the standard hardware installation docs now, you can set the input impedence based off the output impedece of the deck and the output path of the amplifyer is stock 75Ohm but can be easily swapped to 50Ohm for the DdD, oscilloscope video and written guideance is in the docs.

There is already a couple full-set examples with the amplifyers deployed on the Decode Team Internet Archive page, I have done about 30 NTSC transfer jobs since the switchover and have had non of the high frequency dropouts that could happen with a bad impedence match like I observed with the Ace Combat 3 tape.

[Brad]

The VHSDecode (BFF) seems to have signal reflexions (text). Caused by an impedance mismatch somewhere?

This is a fair point. Perhaps I should just wait until I can acquire some rewolf amps (ADA4857), rather than waste everyone's time analyzing possibly-flawed samples. [EDIT: They should arrive in a couple weeks, but I expect it'll be over a month before I get one added to a VCR.]

At that point I should also be able to RF-capture from AG-1980 without the heavy crosshatch and low amplitude that occurs now.

I'm surprised you want to disable TBC, but sure, the next comparison will be Dubbing mode minus 629TBC.

I was not aware that "Dubbing" disables TBC, I though it behaves like "Edit" for JVC models.

Wait, what? Here is the full context where you said you wanted the TBC to be disabled:

I also believe the VCR should be in dubbing mode to disable picture processing and gain more sharpness, I also prefer the 629TBC to be off since not all VCRs have it.

I'm surprised you want to disable TBC, but sure, the next comparison will be Dubbing mode minus 629TBC.

The Image Quality setting (Dubbing or any other) is indeed completely separate from 629TBC. And unlike JVC DigiPure, 3DNR is completely separate from everything else too.

I started to think that the RF hardware is not up to the task, not the decode part of it, Judging by the horizontal resolution of the white text on the blue background, it seems to me that there is not enough bandwidth to capture more details of each scan line.

Domesday Duplicator can capture from LaserDisc and Hi8, both with much higher horizontal resolution than seen here. (I'm not mentioning S-VHS, because the filtering parameters in vhs-decode continue to be tweaked. I don't know what those decodes currently look like.)


(CX cards and other hardware can too, but I mention DdD since that is what I used here.)

[dellsam34]

That's another caveat, So the 629TBC is the main TBC? I was under the impression that it's an extra feature enabled by menu in addition to the regular TBC ON/OFF button on the front Panel, that's what I meant by not all VCRs have it, the VCRs with normal TBC. Appologies for not knowing how that VCR works.

[Brad]

Oh, I see. No apology necessary, sir.

Yes, it is the only TBC in this VCR. I previously showed its front panel button. It's not where you might expect.

On the HS-HD2000U, the light-up green button on the right enables "DigiPure-style" DNR&TBC. On the HV-BS890, that same button is dedicated to the Ghost Reduction Tuner function.





ゴースト
リダクション = gōsuto ridakushon

[lordsmurf]

Yes, it is the only TBC in this VCR.

To me, the inclusion of a semi-exotic VCR is still not showcasing the inabilities/abilities of -decode vs. normal capture setup. It's good, yes, but we all known the gold standard is JVC/Panasonic.

Can you also provide a comparison using "standard" material and not animation? Thanks as always!

Let's go further. I want to see how it performs crowd scenes at events, such as football/soccer, baseball, wrestling (classic WWF), car racing, even golf. That sort of footage is both common, and quite complex and challenging in its own way. That footage has a way of "separating wheat from chaff". I can say right now that ringing/halo is going to obliterate such footage. Whatever "sharpness" is gained will be negligible.

I'd post a response video of a traditional capture of the same scene

The -decode project has weaponized the term "traditional", which never really fit anyway. There's no tradition there. It's just a normal, standard, basic setup. It's not grandma's old pie recipe, but modern archiving tools/methods. (FYI, advanced setups do what -decode still cannot, because it requires live interventions. Even some of the normal gear cannot. So VCR>TBC>card is just basics.)

[dellsam34]

I think the best approach is to take a JVC RF and TBC boards and reverse engineer them to come up with a better modern alternative, This whole computer coding stuff is not cutting it, it's just too many variables, too many ideas, too many hardware to deal with.

If one can start from the heads' rotary transformer coils and work their way to the TBC buffer digital scan lines data, hardware standardization becomes easy, and while at it, get with folks with motor control skills, so the capstan and drum motor can be driven from within the system to ditch the VCR electronics completly, the benifit is any VCR should be able to do any color format and native frame rate and you only need a working mechanism.